non-random ramble

The other day, Alistair chatted to me about the ‘retro’ concept of changing my IDE background to black like in the old skool days. Then I saw a Neal Ford’s blog which also has a black background and looks cool. So I am trying it on this blog.

I think I likes it.

It seems theres a bit of a thing happening out there, a lot of people say black on white is better:

hanselman

coding horror

456bereastreet - really doesnt like it and say you should offer alternative stylesheets.

A lengthy discussion - Makes a good point that if the whole web is black on white and then you come to a site like this, its going to be harder to read.

How do I make sure I am reading enough, or have a good set of working principles to be able to apply to my software design / coding ?

Most people come to this by experience. Experience and interaction with other software developers. Ideas (or memes) spread by word of mouth, more specifically the internet.

Something the scientific community often publishes are review papers. These take a particular field and summarise the findings, providing a massive list of references at the end.

These are extremely useful if you want to then work in that area. Because it has been peer reviewed, there is a certain amount of confidence that crucial works (at least until the date of publication) have been included.

We almost have this in software, but people tend to write books, or blogs about things. These tend to be written by people who have gained a reputation in the field for collating ideas, for example Martin Fowler. Maybe its the same thing, and you get a word of mouth gain in respect for these people.

Perhaps a more formal peer-reviewed process could be realised for software design principles. As a local alternative, I have created a page called ‘Essential Reading’ where I am going to begin to collate such references. Works which I believe contain valuable principles, and which everyone should be aware. I’ll see if it is possible to get something like a review paper going.

I think the goal is that you should be able to communicate about any of these concepts amongst developers and they should be able to understand what your talking about. Its a bit like patterns, but with a wider context, they are principles.

I was inspired to write this after finding Neal Ford’s book which is a book that I had thinking would be great to write. Neal has been working on it for a couple of years and I am looking forward to reading it.

Non-randomly enough, I just found this from Joel which is all about Unicode, but is entitled “The Absolute Minimum Every Software Developer Absolutely, Positively Must Know About Unicode and Character Sets (No Excuses!)” maybe what I’m looking for is a similar article about design principles. Infact, Joel has a list of articles by job role which he has written which is close to what I’m looking for. Also, he co-founded Stack Overflow which is a way forward.

I would also like to start ‘canigrokit.org’ which would be a similar idea but allows you to post snippets of code and people can vote on wether they are readable or not. I have the domain name already.

I just read this interesting response from Brian Eno about the future.

It coincides with some thoughts I’ve had about the future and global warming (I have an imaginary book called ‘Climate of fear’ which I might write).

A while ago, I was performing a refactor in our codebase, pulling some functionality into a core domain object and then factoring it into some smaller, more focused objects.

My main drivers for making this change were:

1) Law of demeter

2) Hard to see all the paths of execution through the method

3) Nested if { ... } statements

Thinking about 2), a method which has an execution path which is difficult to understand immediately and intuitively (for normal beings that is) is likely to hide a bug. Infact this particular piece of code was doing just that which is why I was changing it.

So more complex logic leads to buggy code which leads to a higher cost of maintennance. The latter has a direct impact on the bottom line of your project, not to mention potential success of the software in its usability or even core function.

The refactor followed three steps.

1) Move the method into the Target class. This involved extracting a parameter which was really the knowledge of the client of Traget. I was a little concerned about Target knowing too much which leads to 2). I did this first by moving the method and then running all tests for client class, then I duplicated the test methods on the Target class and ran them (required a small refactor which made them easier to read as a side effect)

2) As i was doing this move, i realised that a scenario had not been tested, a consequence of 2) above, i.e. the path of execution was complex and so a test had been missed (if TDD had been followed this would be unlikely)

3) Finally, I wanted to get rid of the if statements. In order to do this, I used Replace method with method object to pull out the logic inside the if blocks.

Summary :

By decomposing the flow of logic into smaller more focused units, it makes the individual parts easier to understand and so it is more obvious if an error in the flow exists. It also helps to write test cases in a more modular way. Rather than 20 test methods in a test case you break into several test cases covering specific functionality.

One question arises, is refactor 3) a refactor too far ?

How do I decide this ?

The best way is to get someone else to look at the code and see if it reads well to them. If i was writing a book, I would get someone to proof read it so that I could confirm that my language was expressed in a clear way. This is the kind of review which Pair programming provides as standard, but even then it might be useful to show it to people outside the pair as you can both become entangled.

I have been working with Clive on a fair amount of refactoring the last couple of weeks.

The other day we encountered a classic case of refactoring choice and narrowly averted a potentially lengthy (and confusing to future developers) refactor. This was due to my pairs’ breaking the spell of the refactor path we were following.

Our problem was a circular dependancy. We were introducing a new service to the system to replace a heap of code sprinkled throughout the codebase with a central point of access. We had implemented our shiny new service and were moving around the codebase getting things to use it instead of doing all the work for themselves.

We reached a class and attempted to add a reference to our new module (we are working in C#). Doh! visual studio complains at us “you cant add that reference because you already depend on this module”. Apparently there are ways to circumvent this but I am quite a stickler for dependency management (think it was hammered in to me learning C++), and anyway who wants to mess with visual studio when its angry?

Solution 1

Ok, so lets move our dependancy into a module which is higher up the dependancy tree, problem solved. Well, yes, except that once we moved one, we discover a couple more. Plus we are working on a branch and in a codebase without too much of a safety net (ok lets not get into why we would be doing that), so we dont want to be changing the world.

Our Solution was move the class up into a new module, but leave the old one there, just delegating to the new implementation. Not ideal, but at least it leaves the code in an unambigous state (i.e. there are not 2 classes doing the same thing, infact with the same code.

We had reached a point where we were beginning to feel uncomfotable with this. It was feeling heavyweight and intrusive, and worse, felt like we were creating technical debt. this situation would really need to be tidied up later.

We broke for lunch, in the hope that we may have some inspiration with food. After lunch, I returned to find Clive browsing the codebase:

“Hey Clive, did you get any genius ideas over lunch ?”

“Well, I was just looking through the code, and i was wondering, what was it we were trying to achieve again ?”

It was like smelling salts:

“Of course, that’s great problem solving skills, we’ve been going about this all the wrong way!”

The answer was obvious. Instead of moving half the code all over the place because of this one problem child, we had to change that child to be less greedy for responsibility. It was doing too much.

All we had to do was stop it getting its own resources, and instead, pass them in, injecting through the constructor. This pushed the responsibility for obtaining the resource to the client.

As it happened, the client was also in the common module, but was a couple of static methods. A simple move method refactor moved these methods down into a module which we didnt depend upon and hey presto our dependancy circle was broken.

I think this is a great example of how you can get sucked in to a solution. To begin with we found it hard to see another way out of the problem. Simply stopping, looking at the problem form the beginning again, and listening to the code which was telling us not to cross the road, we were able to walk up the street for a few metres and find a safe zebra crossing.